Energy-efficient and secure federated learning framework for wireless sensor networks in smart agriculture applications

The use of wireless sensor networks (WSNs) in smart agriculture has greatly improved data-driven practices such as crop monitoring, irrigation management, and environmental analysis. The deployment of large number of sensor nodes poses challenge for energy, data privacy, and secure communication. In this study, an energy-efficient and secure federated learning (FL) framework is proposed for the smart agriculture application. The proposed approach enables the energy-saving ability of the sensor nodes to train models locally and transmit only learned updates, avoiding the need for raw data transmission as with current centralized learning methods and preserving delicate agricultural information. An additional layer of security is added by implementing a lightweight encryption scheme to secure communications between nodes and the central server. The framework also proposed an energy-aware node selection strategy, which helps to prolong network lifetime by involving only the proper sensor in the learning process. The experimental results on a simulated smart agriculture environment demonstrate the superior performances of the proposed framework in terms of energy efficiency, learning convergence speed, and data protection. In summary, this work is beneficial for the advancement of sustainable, secure and intelligent agricultural systems.